rupture parameters of the 2003 zemmouri (m 6.8), algeria...

Rupture parameters of the 2003 Zemmouri (Mw 6.8), Algeria,

earthquake from joint inversion of interferometric synthetic

aperture radar, coastal uplift, and GPS

Samir Belabbes,1 Charles Wicks,2 Ziyadin Cakir,3 and Mustapha Meghraoui1

Received 3 July 2008; revised 6 December 2008; accepted 12 January 2009; published 20 March 2009.

[1] We study the surface deformation associated with the 21 May 2003 (Mw = 6.8)Zemmouri (Algeria) earthquake, the strongest seismic event felt in the Algiers regionsince 1716. The thrust earthquake mechanism and related surface deformation revealedan average 0.50 m coastal uplift along �55-km-long coastline. We obtain coseismicinterferograms using Envisat advanced synthetic aperture radar (ASAR) (IS2) andRADARSAT standard beam (ST4) data from both the ascending and descending orbits ofEnvisat satellite, whereas the RADARSAT data proved useful only in the descendingmode. While the two RADARSAT interferograms cover the earthquake area, Envisat datacover only the western half of the rupture zone. Although the interferometric syntheticaperture radar (InSAR) coherence in the epicenter area is poor, deformation fringes areobserved along the coast in different patches. In the Boumerdes area, the maximumcoseismic deformation is indicated by the high gradient of fringes visible in allinterferograms in agreement with field measurements (tape, differential GPS, leveling,and GPS). To constrain the earthquake rupture parameters, we model the interferogramsand uplift measurements using elastic dislocations on triangular fault patches in anelastic and homogeneous half-space. We invert the coseismic slip using first, a planarsurface and second, a curved fault, both constructed from triangular elements usingPoly3Dinv program that uses a damped least square minimization. The best fit ofInSAR, coastal uplift, and GPS data corresponds to a 65-km-long fault rupture dipping40� to 50� SE, located at 8 to 13 km offshore with a change in strike west of Boumerdesfrom N60�–65� to N95�–105�. The inferred rupture geometry at depth correlates wellwith the seismological results and may have critical implications for the seismichazard assessment of the Algiers region.

Citation: Belabbes, S., C. Wicks, Z. Cakir, and M. Meghraoui (2009), Rupture parameters of the 2003 Zemmouri (Mw 6.8), Algeria,

earthquake from joint inversion of interferometric synthetic aperture radar, coastal uplift, and GPS, J. Geophys. Res., 114, B03406,

doi:10.1029/2008JB005912.

1. Introduction

[2] The thrust and fold system of the Tell Atlas (northernAlgeria) has been the site of several large and moderateearthquakes in the last decades (Table 1). This shallowseismic activity was very often associated with surfacefaulting, the conspicuous example being the El Asnamthrust faulting and related 1980, Mw 7.3 major event [Philipand Meghraoui, 1983; Yielding et al., 1989]. The thrustmechanism (global centroid moment tensor (CMT)) of theMw 6.8, 2003 Zemmouri earthquake which occurred in theTell Atlas tectonic belt, confirms the pattern of activedeformation that predicts 4–6 mm/a of convergence along

the Africa-Eurasia plate boundary (Figure 1a) [Meghraouiet al., 1996; Nocquet and Calais, 2004; Serpelloni et al.,2007]. Its coastal location hindered, however, the directobservation of surface ruptures and has left open questionson the probable fault scarp location, structural character-istics and geometry of the causative fault [Meghraoui et al.,2004; Deverchere et al., 2005].[3] The 2003 Zemmouri earthquake (also called some-

times the Boumerdes earthquake) affected the eastern edge ofthe Mitidja Quaternary basin (Figure 1b), and was respon-sible for severe damage �20 km east of the capital cityAlgiers [Ayadi et al., 2003; Harbi et al., 2007]. The NE-SW striking fault is consistent with the Tell Atlas fold andthrust tectonics, the uplifted 55-km-long coastal shoreline[Meghraoui et al., 2004], the �40� SE dipping aftershocks[Ayadi et al., 2008] and the inversion of body waves[Delouis et al., 2004; Yagi, 2003]. A detailed bathymetricsurvey and seismic profiles offshore describe the morpho-logical and structural features in the Mediterranean Sea[Domzig et al., 2006]. A downdip model of uniform slip

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114, B03406, doi:10.1029/2008JB005912, 2009ClickHere

for

FullArticle

1Institut de Physique du Globe de Strasbourg, UMR 7516, Strasbourg,France.

2U.S. Geological Survey, Menlo Park, California, USA.3Faculty of Mines, Istanbul Technical University, Istanbul, Turkey.

Copyright 2009 by the American Geophysical Union.0148-0227/09/2008JB005912$09.00

B03406 1 of 16

http://dx.doi.org/10.1029/2008JB005912

on a rectangular dislocation deduced from direct measure-ments of coastal uplift (using differential GPS (DGPS) andconventional leveling [Meghraoui et al., 2004]) is consis-tent with the slip distribution at depth inferred from theinversion of seismic and geodetic data [Delouis et al.,2004]. These models of seismic source include surface slipand comply with the observed tsunami effects and relatedtide gauge records in the western Mediterranean Sea [Alassetet al., 2006].[4] The analysis of SAR images and related interfero-

grams in the last decade has led to a significant progress inthe understanding of earthquake deformation [Massonnetand Feigl, 1998; Burgmann et al., 2000; Wright et al.,2004]. The phase interference of spaceborne radar imageshas the outstanding advantage to locate and display the fielddisplacement of large sections of any earthquake area,providing that an appropriate coherence level exists be-tween SAR images. The generation of SAR interferogramsis therefore particularly important in thrust and fold systemswith complex surface deformation (seismic and aseismic)[Fielding et al., 2004] as it may lead to a better seismichazard assessment. The application of interferometric syn-thetic aperture radar (InSAR) in active zones of NorthAfrica has been recently performed for the Mw 5.7, 1999Ain Temouchent blind thrust earthquake [Belabbes et al.,

2008] and for the Mw 6.0, 1994 and Mw 6.4, 2004 AlHoceima earthquakes which revealed the existence of blindand conjugate seismic ruptures in the Rif Mountain belt[Cakir et al., 2006; Akoglu et al., 2006] (Figure 1a). Thecoastal location of the Zemmouri earthquake is not an idealconfiguration for producing interferograms since most ofthe earthquake deformation occurred offshore. Neverthe-less, the impressive coastal uplift and epicenter locationimply that significant surface deformation took place inlandas well and can be studied using SAR images.[5] In this paper, we first present the seismotectonic

characteristics of the Zemmouri earthquake and provide adetailed study of interferograms obtained from RADAR-SAT and Envisat images. The InSAR data document themeasurements of displacement field in the earthquake areain agreement with previous studies of coastal uplift. Mod-eling of the InSAR data set together with the uplift and GPSmeasurements allow us to deduce the fault parameters andassociated coseismic slip distribution. Finally, open ques-tions on the earthquake fault geometry and the importanceof InSAR analysis are discussed emphasizing the constraintof a hidden seismogenic thrust rupture with coastal geodetic,tectonic and seismologic data.

2. Seismotectonic Setting

[6] The Tell Atlas of northern Algeria experienced thelargest recorded earthquake at El Asnam (10 October 1980,Mw 7.3) related to a NE-SW trending reverse fault [Ouyed etal., 1981; Philip and Meghraoui, 1983]. Other recent largeand moderate seismic events of this region (Figures 1a and1b and Table 1), including the Zemmouri earthquake of2003 revealed a comparable pattern of active deformation inagreement with the NNW-SSE to NW-SE convergence andtranspressive tectonics along the Africa-Eurasia plateboundary [Meghraoui et al., 1996; Stich et al., 2003]. Theseseismogenic structures of the Tell Atlas are related to a

Table 1. Large and Moderate Earthquakes With Thrust Mechan-

ism Along the Tell Atlas

Location Date Longitude (deg) Latitude (deg) Mw

Orleansville 9 Sep 1954 1.47 36.28 6.7El Asnam 10 Oct 1980 1.36 36.18 7.3Tipaza 29 Oct 1989 2.92 36.84 5.9Mascara 18 Aug 1994 �0.03 35.40 5.7Ain Temouchent 22 Dec 1999 �1.45 35.34 5.7Beni Ourtilane 10 Nov 2000 4.69 36.71 5.7Zemmouri 21 May 2003 3.65 36.83 6.8

Figure 1a. Shaded relief map of eastern Mediterranean with focal mechanism solutions of earthquakesbetween 1980 and 2005 (data from Global CMT catalog). Gray mechanism corresponds to the 2003Zemmouri earthquake. The gray line represents the plate boundary according to Meghraoui et al. [1996].Black arrows are direction of convergence, and the number is plate velocity in mm/a [Nocquet andCalais, 2004]. The squared area is for Figure 1b.

B03406 BELABBES ET AL.: INSAR OF THE ZEMMOURI EARTHQUAKE

2 of 16

B03406

complex ‘‘en echelon’’ thrust and fold system located alongan E-W trending narrow strip parallel to the coastline. Thelate Quaternary active deformation that indicates a rate of 2to 3 mm/a. for compressive movements estimated from theshortening of folded units and paleoseismic investigations[Meghraoui and Doumaz, 1996], is comparable to the 4 to6 mm/a. convergence rate obtained from NUVEL-1A andglobal GPS solutions along the plate boundary [Nocquetand Calais, 2004; Serpelloni et al., 2007].[7] The 2003 Zemmouri earthquake affected �55-km-

long coastline and �15-km-thick uppermost crustal struc-ture that belongs to the eastern regions of the Mitidja Basinand related Blida thrust and fold system [Ayadi et al., 2008].A detailed study of the main shock relocation [Bounif et al.,2004] and aftershock distribution (that covers 2 months)using the tomography analysis reveals a NE-SW trendingand 40�SE dipping fault zone with two seismicity patches[Ayadi et al., 2008]. The double difference seismic analysisshows a large concentration of aftershocks and depthdistribution SW of the fault rupture and indicates a SEdipping thrust geometry in agreement with the centroidmoment tensor (CMT) solution (Figure 2 and Table 2)[Ayadi et al., 2008]. Using an analysis of body waves andsurface waves of teleseismic records, Delouis et al. [2004]found 2.86 � 1019 N m seismic moment and a �15 s fault

rupture with �1 m surface slip, and 2.1 m maximum slip atdepth with bilateral rupture propagation along two patches(Table 2). Yagi [2003] also constructed source models frombody wave inversion and obtained similar results pointingout the bilateral rupture propagation on a 60-km-long faultwith 2 patches and 2.3 m maximum slip at depth (Table 2).The inversion of geodetic data (coastal uplift and GPS) andaccelerograms also points out the two patches with up to 1 msurface slip SW of the earthquake rupture [Semmane et al.,2005]. According to the fault geometry, seismic moment,main shock location at depth and field observations, theearthquake fault did not rupture the surface along thecoastline but likely offshore at the seafloor. Detailed fieldinvestigations made immediately after the main shock indi-cate, however, the existence of 2- to 3-km-long N95�–100�trending surface cracks observed along the Thenia fault(Figure 1b).[8] Coseismic uplift of marine terraces along the 55-km-

long coastline (Figure 3) combined with GPS and conven-tional leveling indicate an average of 0.55 m verticalmovement with two slip patches from which a SE dippingplanar dislocation model is resolved with 2.8 � 1019 N mgeodetic moment [Meghraoui et al., 2004]. Although basedon a 60-km-long and 15-km-wide simple rectangular dislo-cation, the model suggests a seismogenic thrust rupture

Figure 1b. Envisat/RADARSAT radar frames (dashed rectangles) with arrows indicating the satelliteflight direction for ascending and descending orbits of the 21 May 2003 Zemmouri earthquake area. Red,black, and blue arrows indicate flight direction for satellites. The red star indicates the earthquakeepicenter location [Bounif et al., 2004], and focal mechanisms are from the Global CMT solution between1980 and 2003 (size of beach balls is according to magnitude and red mechanism is for the Zemmouriearthquake). Black lines indicate active thrust faults, and dashed arrowed lines indicate active folding andwest of the 2003 epicenter TF is for Thenia Fault [Meghraoui, 1988].


3 of 16

B03406

dipping 50� SE consistent with the seismological results(Table 2). The coastal deformation and rupture historyassociated with the aftershock distribution (Figure 2), show-ing a clear concentration of seismic events on the SW patchsuggest, however, a complex fault rupture likely made ofseveral asperities along strike. Early results from geodeticmeasurements (coastal uplift and GPS) and seismicityanalysis [Bounif et al., 2004; Meghraoui et al., 2004;Delouis et al., 2004; Yelles et al., 2004; Semmane et al.,2005; Alasset et al., 2006] imply the existence of a N54�–N70� striking and 40� to 50� SE dipping fault plane withsurface slip at 5 to 15 km offshore. From the high-resolutionswath bathymetry and seismic profiles Deverchere et al.[2005] study the seafloor of the continental slope offshorethe 2003 earthquake area, identify outcropping thrust faultand scarps located at �20 and 32 km from the shoreline andinfer a flat and ramp rupture geometry. The moment tensoranalysis obtained from broadband seismic records yields a

25� SE dipping rupture (Table 2) that suggests a faultsurface trace at 15 to 20 km offshore limited to the SWby a NNE trending transform fault and forms a step overwith the Blida thrust system [Braunmiller and Bernardi,2005].

3. The InSAR Analysis

[9] We have calculated four coseismic interferograms,one ascending and three descending, using synthetic apertureradar (SAR) data acquired by the Canadian Space Agency’sRADARSAT satellite and European Space Agency’s Envisatsatellite (Table 3). The raw data were processed using thecommercial GAMMA software with 5 azimuth 1 range look(i.e., averaged to 20� 20 m of ground pixel size) and filteredusing weighted power spectrum technique of Goldstein andWerner [1998]. The effect of topography that depends onthe perpendicular separation between orbital trajectories is

Figure 2. The 21 May 2003 Zemmouri earthquake (black star, Mw 6.8) and aftershock distribution(white circles) of seismic events from 25 May to 31 July 2003 [Ayadi et al., 2008]. The main shock andmain aftershocks (black circles) are relocated using the double difference seismic analysis of Bounif et al.[2004]. Focal mechanism solutions are Global CMT.

Table 2. Fault Plane Parameters of the 21 May 2003 Zemmouri Earthquake

SourceLongitude(deg)

Latitude(deg)

Depth(km)

Plane 1a Plane 2

M0 (N m)Strike(deg)

Dip(deg)

Rake(deg)

Strike(deg)

Dip(deg)

Rake(deg)

HRV 3.58 36.93 15 57 44 71 262 49 107 2.01 � 1019

USGS 3.78 36.89 9 54 47 88 237 43 92 1.30 � 1019

INGV 3.61 36.9 15 65 27 86 250 63 92 1.80 � 1019

Yagi [2003] 3.65b 36.83b - 54 47 86 - - - 2.40 � 1019

Delouis et al. [2004] 3.65b 36.83b - 70 40 95 - - - 2.80 � 1019

Meghraoui et al. [2004] 3.65b 36.83b - 54 50 88 - - - 2.75 � 1019

Braunmiller and Bernardi [2005] 3.65b 36.83b - 62 25 82 - - - 3.48 � 1019

This study planar model 3.65b 36.83b 8 65 40 90 - - - 1.78 � 1019

This study curved model 3.65b 36.83b 10 65 40 90 - - - 2.15 � 1019

aPlane 1 is the principal fault according to field observations.bEpicenter relocation by Bounif et al. [2004].


4 of 16

B03406

removed from the interferograms using the SRTM 3-arc-sec(�90 m) posting digital elevation model [Farr et al., 2007].Envisat interferograms were obtained using precise orbitsfrom Delft University [Scharroo and Visser, 1998] while forRADARSAT interferograms orbital information are fromthe data header.[10] Baselines for Envisat data were not reestimated

because there are no orbital residuals visible farther southof the earthquake area where there is supposedly no surfacedeformation. The available RADARSAT orbits are impre-cise and processing them leaves orbital errors. Therefore,we have modeled the orbital residuals with a 2-D quadraticsurface and subtracted it from the interferograms [Zebker etal., 1994].[11] While the descending RADARSAT frame covers

entirely the uplifted coastal region from Cap Matifou tothe west and Tighzirt to the east, both the ascending anddescending Envisat frames cover only the western part ofthe earthquake area (Figure 1b). Despite the atmosphericnoise and low coherence particularly in the epicentral area,the interferograms display clear coseismic fringes all alongthe coast (Figure 4). The poor coherence in the epicentral

region exists in all the interferograms and is most likely dueto the high agricultural activity along the Oued Isser River.Signal decorrelation in the Envisat interferograms also occursto the south in the mountainous regions due to the lowaltitude of ambiguity of the interferometric pairs (Table 3).Because of the significant change in the look direction(Table 4), the interferograms naturally differ from eachother. However, the difference is not remarkable since thedeformation is overwhelmingly vertical, which can also beseen in Figures 4e, 4f, and 4g that show digitized fringes ofall interferograms. The difference in the fringe patternbetween the interferograms is mostly due to the horizontalcomponent of the surface displacement. RADARSATinterferograms indicate that there are two lobes of defor-mation centered in the Boumerdes and Cap Djenet regionswith the earthquake epicenter being located in the area oflower deformation in between them (Figures 4a and 4b).All interferograms indicate that the maximum surfacedeformation occurred indeed in the Zemmouri-Boumerdesregion, confirming the field measurement of the coastaluplift [Meghraoui et al., 2004]. In this region, up to 14fringes can be counted in the RADARSAT interferograms

Figure 3. Coastal uplift at 6 km west of Dellys (Figure 1b). The photograph shows the elevated marineterrace and platform (black arrow is the former shoreline and white arrow is the new shoreline) with anaverage 0.55 m of uplift [Meghraoui et al., 2004].

Table 3. Characteristics of SAR Images Covering the 2003 Zemmouri Earthquake Area

Satellite Beam Mode Master Orbit Master Date Slave Orbit Slave Date TrajectoryAltitude of

Ambiguity (m) B perp (m)

RADARSAT1 ST4 35996 27 Sep 2002 40798 29 Aug 2003 Descending 1358 12RADARSAT1 ST4 35996 27 Sep 2002 41141 22 Sep 2003 Descending 84 194Envisat IS2 4900 6 Feb 2003 6904 26 Jun 2003 Descending 22 238Envisat IS2 5308 6 Mar 2003 7312 24 Jul 2003 Ascending 40 419


5 of 16

B03406

Figure 4


6 of 16

B03406

(Figures 4a and 4b) and 16 fringes in the Envisat interfero-grams (Figures 4c and 4d). Assuming purely verticaldeformation, these line-of-sight (LOS) displacements corre-spond approximately to 0.50 m of uplift since the Envisatinterferograms are more sensitive to vertical deformationthan the RADARSAT interferograms. Captured only by theRADARSAT interferograms, the LOS displacement in theCap Djenet region reaches to �0.25 m (i.e., 9 fringes).[12] An examination of interferograms reveals some

anomalies in the fringe pattern of the Cap Matifou region(black arrows in Figure 4). Fringes are clearly disturbed,offset and inverted along a lineament that coincides with theThenia fault. This implies that the Thenia fault must haveexperienced some triggered slip during or after the mainshock, a phenomenon commonly observed due to largeearthquakes [Fialko et al., 2002].

4. Modeling the InSAR Data

[13] We used the Poly3Dinv slip inversion method tomodel geodetic data using realistic fault surfaces (see alsomaps in Figures 5a and 5b) [Thomas, 1993; Maerten et al.,2005]. The method is based on the analytical solution for atriangular dislocation in a linear, elastic, homogeneous andisotropic half-space, which uses triangular surfaces as dis-continuities [Maerten et al., 2005]. Hence, the use oftriangular elements allowed us to construct fault modelsthat better approximate two-dimensional planar surfaces,avoiding gaps and overlaps that are inevitably encounteredwhen modeling highly segmented faults of varying strikewith simple rectangular dislocations. This method improvesthe fit to the geodetic data particularly in the near field whenmodeling complicated fault ruptures [Maerten et al., 2005;Resor et al., 2005]. Fault surfaces meshed with triangleswere constructed using MATLAB1. We meshed the planaroffshore fault with 6 � 5 quadrangles (i.e., 60 triangles) and2 � 1 quadrangles for the Thenia fault. The slip distributionof the triangle elements was then inverted for, with anegativity constraint on the dip slip component (i.e., thrustonly, 90� rake). To avoid unphysical oscillatory slip, thescale-dependent umbrella smoothing operator of Poly3Dinvis applied to the inverted slip distribution. We choose thesmoothing factor based on the RMS misfit versus roughness

curve; this selected factor (i.e., 0.4) represents the bestcompromise between the roughness of slip and misfit tothe data (Figures 5a and 5b). In our inversions, we usedigitized fringes because some of them cannot be unwrappeddue to noise and unwrapping errors. In this case, the residualinterferograms are obtained by a multiplication in thecomplex domain of the data and the synthetic interferograms.On the basis of the fringe gradient, the southernmost fringein the isolated patch located near Cap Djenet (Figure 4e) isassumed to be the third fringe (i.e., 8.49 cm) and the largecurved fringe southeast of Boumerdes (Figure 4f) is as-sumed to be the first fringe (i.e., 2.83 cm). A constant offsetin the InSAR data is not solved as we think it is negligiblesince one may observe to the south areas of nominaldeformation.[14] In order to constrain the location and geometry of the

fault rupture, we run several inversions using digitizedfringes (Figures 4e, 4f, and 4g), coastal uplift measurements[Meghraoui et al., 2004] and coseismic GPS vectors [Yelleset al., 2004]. The InSAR, coastal uplift and GPS data weretreated as equally weighted in all inversions. The Poissonratio in the elastic half-space is given as 0.25. A tilt fororbital residuals is solved by the Poly3D inversion butfound to be insignificant. Several attempts with no con-straints on the fault rake are performed but the inversionpredicts abnormally high strike-slip (both right and leftlateral) components. Therefore, we keep the N60�–65�strike and rake as pure thrust (90�) fixed and test severalfault plane dips (20�–60�) in agreement with most of focalmechanism parameters of the 2003 main shock (Table 2). Inthe absence of any observed seafloor rupture and relatedSAR data, no surface slip on the uppermost patches isallowed in the inversions. However, it is possible thatsome slip reached the seafloor, but neither the SAR datanor the uplift measurements can resolve it. Taking intoaccount the surface deformation and seismotectonic frame-work of the epicentral area, we suggest the following twoslip models:[15] 1. A variable slip model is obtained using inversions

with various planar faults striking N65� and dipping 30�,40�, and 50� SE located from 6 km to 24 km offshore(Figure 5a). The best fitting model (RMS = 2.69 cm) isobtained with 60-km-long and 30-km-wide offshore thrustfault (i.e., fault 4 in Figure 5a) striking N65�, dipping 30� to40� SE and located at �13 km offshore from the epicenter(Figures 5a and 6a). However, as illustrated in the RMSversus distance plot of Figure 5a, the fault tip location is nottightly constrained and could be anywhere between 9 to18 km from the shoreline. The best fitting model predicts acoseismic slip distribution at �8- to 10-km-depth on the

Figure 4. Coseismic interferograms of the western deformation zone of the 2003 Zemmouri earthquake area in (a and b)the descending RADARSAT geometry and (c and d) ascending and descending Envisat radar geometry; the imagereference (date and orbit numbers) is in the top right corner. Black arrows indicate the satellite flight direction. Star showsthe 2003 Zemmouri earthquake epicenter [Bounif et al., 2004]. (e, f, and g) Digitized fringes labeled with displacement (incm) in the line-of-sight (LOS) direction for each interferogram (Figure 4e with 4b, Figure 4f with 4c, and Figure 4g with4d). (h) baselines with altitudes of ambiguity in m (in black box) for each pair of SAR images with orbit numbers. Fringesare observed in all interferograms with 2 lobes of high deformation centered in Boumerdes and Cap Djenet region. N105�trending black arrows near Cap Matifou show the anomalies observed in the fringe pattern along the Thenia Fault(Figure 1b). The western earthquake area between Cap Matifou and Zemmouri shows several fringes that correspond to thezone of maximum deformation.

Table 4. Unit Vectors of SAR Scenes

Satellite Flight Direction East North Up

RADARSAT Descending 0.59339 �0.10980 0.79739Envisat Descending 0.37188 �0.07946 0.92487Envisat Ascending �0.38526 �0.08173 0.91918

7 of 16

B03406 BELABBES ET AL.: INSAR OF THE ZEMMOURI EARTHQUAKE B03406

fault and shows two linked asperities reaching 4.7 m and�3 m maximum slip in the NE and SW patches, respec-tively, and 1.78 � 1019 N m (Mw 6.8) geodetic moment.Additionally, offset fringes observed along the Thenia Fault

(TF in Figure 1b) in Cap Matifou are modeled in a forwardmanner with a 0.38 m slip on a �20-km-long subverticalright-lateral fault as the inversions fail to explain thedisturbed fringes (Figure 4).

Figure 5


8 of 16

B03406

[16] 2. The alternative model takes into account the com-plicated fringe distribution west of Boumerdes (Figures 4a,4b, 4c, and 4d) and also includes the inversion of coseismicslip on regularly spaced faults (�2 km) from 4 to 20 kmoffshore (Figure 5b). To the NE, faults are planar with anazimuth of N60–65�, consistent with the aftershocks dis-tribution [Ayadi et al., 2008] and focal mechanism solutionsof the main shock (Table 2). The SW fault section ismodeled according to the N95�–105� trending aftershocksand related tomography [Ayadi et al., 2008] where the mainNE-SW trending earthquake rupture cannot crosscut thecontinuous fringes visible particularly in the ascendingEnvisat interferogram west of Boumerdes (Figure 4c). Thischange in fault strike may be modeled as a curved fault. Theslip on each fault is inverted for dips 20�, 30�, 40�, 50� and60� SE using the same InSAR, GPS, and coastal uplift data.As shown in Figure 5b, the inversion results suggest that theminimum RMS misfits improve the fault location between8 and 13 km offshore. The lowest 2.67 cm RMS errorindicates that the fault rupture is most probably located

between 8 and 9 km offshore (from the epicenter) for a faultdipping 40� and 50� SE. The RMS misfit plot also showsminimum values (2.77 cm, 2.81 cm) for 30� SE dippingfault at 11 km offshore and for 20�, 30�, 40�, and 50� at 13 kmoffshore (Figure 5b). Taking into account the 40� to 50� SEfault dip of most focal mechanism solutions (Table 2) andthe 40� SE dipping fault geometry from aftershocks tomog-raphy analysis [Ayadi et al., 2008], we select fault 3 inFigure 5b with 40� to 50� SE dip as our preferred solution.The fault is here a 65-km-long and 30-km-wide offshorethrust fault striking N65� and dipping 40� SE, to the nearlyvertical and �N102� trending rupture SW of the epicenter(Figure 5b). Two distinct slip patches yield 2.15 � 1019 N m(Mw 6.8) geodetic moment with �0.5 m slip reaching theseafloor (Figure 6b). Additionally, offset fringes observed inCap Matifou (Figures 4a, 4b, 4c, and 4d) may be explainedby 0.15 m triggered slip on the 20-km-long subverticalright-lateral secondary Thenia fault (Figure 1b).[17] Figure 7 shows the synthetic interferograms pre-

dicted by the best fitting models with planar (Figures 7a–

Figure 5. (a) Map of the different planar fault tip positions (1 to 8) used for the inversions. The white star represents themain shock epicenter [Bounif et al., 2004] and the reference point for distance calculation. Note that fault 4 (white dashedline) represents in this case the preferred fault model taking into account the lowest RMS misfit (right inset). The rightinset shows the RMS error versus distance from the epicenter for different faults (1 to 8) with various dips (30�, 40� and50�). The arrow shows the preferred solution at �13 km. The left inset shows the model roughness versus the RMS misfit,the corresponding smoothing factor for each model roughness is indicated on the right axis. (b) Map of different curvedfault tip positions (dashed lines) used for the inversion with various dips (20�, 30�, 40�, 50� and 60�). The lowest RMSerror is calculated for the 40� and 50� curved fault at �9 km (white dashed line 3) from the epicenter and represents ourpreferred model. The right inset shows the RMS misfit versus the model roughness with 0.4 corresponding smoothingfactor.

Figure 6. (a) Slip distribution model of preferred planar fault (4 in Figure 5a) of the Zemmouriearthquake rupture obtained from the inversion of the digitized fringes and coastal uplift data. Dipcomponent of the coseismic slip on each triangular element are inverted using Poly3Dinv. Dip-slipdistribution (4.7 m maximum from colored contours on fault surfaces) shows two linked maximum ofslip on the N65� trending and 60-km-long major fault rupture. (b) Slip distribution of the preferred curvedfault model striking N60� (�65 km) to N102� (�13 km) on the western part. The curved fault model isconstructed with Poly3D in order to match the two fault planes with different strikes. Dip-slip distribution(2.1 m maximum from colored contours) with two distinct maximum slips reaching the seafloor north ofBoumerdes and Cap Djenet.

9 of 16


Figure 7


10 of 16

B03406

7c) and curved (Figures 7d–7f) faults, together with theobserved fringes (black lines) for comparison. The mainfeatures of the observed interferograms are successfullyreproduced by the best fit models also illustrated by theLOS profiles (Figure 7). The modeled interferograms alsoshow the difference of surface displacement field (syntheticfringes) with respect to the offshore fault location anddimensions. We also note that the inferred horizontaldisplacement at the SW end of the planar fault model(Figure 7a) displays a better fit with the coseismic GPSmeasurements [Yelles et al., 2004]. A comparison betweenvertical movements obtained from field measurements(coastal uplift and leveling) and the inversion modelsindicates that the curved fault model provides a much betterfit (Figures 8a and 8b). On the other hand, the verticaldisplacement predicted by the model in Figures 8a and 8b isin good agreement and within the error range (±0.10 m) ofthe uplift measurements [Meghraoui et al., 2004]. Thecomparison in Figure 8 also shows that the easternmostmodeled uplifts are underestimated for the planar faultmodel. The inferred geodetic moments of 1.78 � 1019 N m(Mw 6.8) and 2.15 � 1019 N m (Mw 6.8) determined fromthe inversion models are, however, comparable and in goodagreement with those obtained from seismological observa-tions (Table 2). The remaining fringes in the residualinterferograms of Figures 9a, 9b, 9c, 9d, 9e, and 9f and inparticular for RADARSAT may result from atmosphericartifacts and unaccounted complexity of the rupture geom-

etry, especially around the western termination of theearthquake rupture where the Blida, Thenia and Zemmourifaults meet (Figure 1b).

5. Discussion and Conclusion: Constraints of aHidden Thrust by InSAR Along a Coastal Area

[18] The crustal deformation associated with the coastal2003 Zemmouri earthquake is documented using the InSARanalysis combined with uplift measurements and GPS data.The surface displacements obtained by the InSAR, coastaluplift and GPS measurements on a maximum of 20-km-wide strip along the coastline helped us to constrain therupture geometry and its location offshore. The fringesobserved on the Envisat and RADARSAT interferogramsof the earthquake area provide an important data set for themodeling of the fault rupture. The two inversion models ofsurface displacements provide geodetic moments of 1.78 �1019 N m (Mw 6.8) and 2.15 � 1019 N m on a 30� to 50� SEdipping fault rupture whose tip is located at �8 to 18 kmoffshore from the coastal epicenter (Figures 5a and 5b). Ourmodeling clearly suggests an offshore N60�–65� trendingfault rupture, dipping 40� to 50� SE and located at about 8 to13 km from the coast (Figures 5 and 8). The seismicmoment, fault geometry and slip distribution are in goodagreement with the main shock and aftershock distributionand teleseismic waveform modeling (Table 2) [Bounif et al.,2004; Delouis et al., 2004; Ayadi et al., 2008].

Figure 7. Synthetic fringes obtained from the inversion of the digitized fringe lines (in black as in Figures 4e, 4f, and 4g)obtained from Envisat ascending and descending, and (a, b, and c) RADARSAT descending for the planar fault and (d, e,and f) curved fault. Dashed rectangle represents the modeled fault area. The fringe pattern is labeled with LOS displacementin cm (white box). Numbered LOS profiles on maps are indicated below. The dotted black line to the west is the Theniafault. The red star is the relocated main shock epicenter [Bounif et al., 2004]. (left) The planar fault model predictshorizontal surface slip (blue arrows) in agreement with GPS measurements (white arrows [Yelles et al., 2004]) except forthe northward vector immediately west of the epicenter. (right) The curved fault model also provides a good fit with theGPS directions, but the amount of measured displacements is larger. The geodetic moment (top left corner) obtained frominversion of InSAR, uplift, and GPS data is consistent with those obtained from seismology (Table 2).

Figure 8. Vertical component of surface displacement (contour lines in cm) predicted by the preferredmodels for the (a) planar and (b) curved faults. Note that the curved fault solution better explains theuplift measurements east of Cap Djenet.


11 of 16

B03406

5.1. InSAR Resolution and Sensitivity

[19] Although atmospheric noise and low coherence inRADARSAT images hampered the quality of interfero-grams, coseismic fringes are still visible by patches in theBoumerdes and Cap Djenet-Dellys areas (Figure 4). Thelack of information on surface deformation in interfero-grams near the epicenter zone and areas of low coherence

led to some difficulties to recognize the nominal deforma-tion. Nevertheless, field investigations of coastal upliftmeasured with tape, leveling and DGPS [Meghraoui etal., 2004] and GPS measurements [Yelles et al., 2004]correlated with the fringe distribution contribute to betterresolve uncertainties of the surface deformation. Althoughthe InSAR data cover a period of �1 year for RADARSAT

Figure 9. Residual interferograms (a, b, and c) for planar fault model and (e, f, and g) for the curvedfault model obtained after subtracting the synthetic interferograms (Figure 7) from the observed data(Figures 4a, 4b, 4c, and 4d). The residuals are negligible for Envisat interferograms but more significantin the case of RADARSAT interferograms probably due to atmospheric artifacts and unaccountedcomplexity of the rupture geometry. Reference of SAR images is in top left corner.

12 of 16


and �4 months for Envisat (Table 2), the interferogramsreflect mainly the coseismic deformation. Indeed, postseis-mic slip obtained from GPS measurements in the earth-quake area indicates �1 cm/a between 2003 and 2006[Mahsas et al., 2008] and do not add a significant displace-ment in interferograms. The locally inverted fringes visiblein RADARSAT interferograms about 10 km WSW ofBoumerdes may be interpreted as a local artifact possiblydue to water pumping for agricultural purposes in theMitidja Quaternary basin. However, the consistent fringedistribution in the SW end of the earthquake rupture zonemay include surface displacement due to the main after-shocks of 27 May (Mw 5.8), 28 May (Mw 4.9), and 29 May(Mw 5.0, Figure 2). In comparison with other case studies ofInSAR applied to small surface deformation [Lohman andSimons, 2005; Belabbes et al., 2008], these moderate andshallow main aftershocks may in fact generate 3 to 5 fringesin addition to the main shock coseismic surface deformation.

5.2. Fault Geometry and Surface Deformation

[20] The integration of InSAR data with coastal uplift andGPS results combined with the main shock and aftershockstudy [Ayadi et al., 2008] constrains the 2003 earthquakefault location, dimension and geometry at depth. The 2003earthquake rupture is hidden by the sea and although the

surface deformation measured inland infers only a partialpicture of the fault rupture, our modeling suggest that thefault rupture may have reached the seafloor �8 to 18 kmnorth of the epicenter (Figures 5a and 5b). However, thelowest RMS errors (<2.7 cm) suggest surface faultingbetween �8 and 13 km offshore from the epicenter on a40� to 50� SE dipping fault rupture (Figures 5 and 10).Using several fault dips, the RMS misfits of Figures 5a and5b show a clear difference between the planar fault model(Figure 5a) and the curved fault model (Figure 5b). Usingthe tomographic analysis of the main shock and aftershocksdistribution, the N60�–65� trending and 40� to 50� SEdipping rupture geometry illustrate the fault plane at depthand related seafloor rupture at a distance <10 km offshore(Figure 11) [Ayadi et al., 2008]. The fixed N54� strike and�15 km distance of the fault rupture from the epicenter ofSemmane et al. [2005] imply that the NE fault section is at ahigher distance (>20 km) and do not conform with thesignificant coastal uplift measured in the Cap Djenet-Dellysregion (Figure 3). The inversion of seismic waveformscombined with geodetic data (coastal uplift and GPS) dis-plays a 40� SE dipping with surface faulting and slip at �7to 9 km distance from the epicenter [Delouis et al., 2004].The forward modeling of coastal uplift measurements infera 50� SE dipping fault plane and locate a possible seafloor

Figure 10. Location and dimension of 2003 Zemmouri earthquake fault rupture from different modelsand authors (see text for explanation). The red star is the relocated epicenter [Bounif et al., 2004], and theaftershock distribution is from Ayadi et al. [2008]. The N105� trending dashed line south of Cap Matifoucorresponds to the nearly vertical Thenia fault. Our planar fault model (blue rectangle) and curved faultmodel (black rectangles, see also Figures 5 and 6) comply with the InSAR, coastal uplift, and GPS dataand take into account the large coverage of InSAR data in the earthquake area.


13 of 16

B03406

rupture at 5 to 10 km from the epicenter [Meghraoui et al.,2004]. Hence, the joint analysis of geodetic and seismolog-ical data constrains the SE dipping fault geometry andrupture location of the 2003 Zemmouri earthquake. Thebathymetric survey coupled with seismic profiles offshore(>15 km from the coastline) the earthquake area, locate twooutcropping thrusts and related scarps at �20 and 32 kmoffshore and infer a flat and ramp fault geometry for the2003 rupture [Deverchere et al., 2005]. In Figure 10, theseafloor scarps located at the base of the continental slopeare scarps B1 and B2 interpreted by Deverchere et al.[2005] as the prolongation of 2003 rupture; however, thisinterpretation that implies a flat and ramp fault geometry(Figure 11) do not take into account the coastal uplift data.If scarps B1 and B2 located at �20 km from the epicenterare directly related to the 2003 rupture, they imply a slippropagating to the surface and a much higher seismicmoment (�Mw 7.3) for the Zemmouri earthquake. In fact,sea bottom rupture scarps at � 15 km distance from thecoastline do not match an Mw 6.8 main shock and after-shock location, related coastal uplift and fault modelsobtained from geodetic data (see fault plane location inFigures 6, 10, and 11).[21] The SE dipping planar and curviplanar thrust fault

models shows two slip patches (Figures 6a and 6b) com-parable to the size of thrust segments in the Tell Atlas[Meghraoui, 1988] that are likely controlled by the tectonicframework of the eastern Mitidja basin (Figure 1b). The

tectonic background determines the thrust length, strike anddip that control the rupture patches and related size of thrustsegments during large and moderate earthquakes. Indeed,surface ruptures of the large El Asnam earthquake (Table 1)displayed 36-km-long thrust fault segment with 6 m max-imum surface slip [Philip and Meghraoui, 1983]. Themaximum slip distribution in the two patches of the Zem-mouri earthquake reaches 4.7 m and 2.3 m for the planarand curved fault models, respectively, with surface slipreaching �0.50 m in the latter. A comparable amount of�1 m surface slip is also obtained by Delouis et al. [2004]and Semmane et al. [2005].

5.3. Thenia Fault Branch

[22] To the SW, the 2003 earthquake rupture propagationsuggests a fault interaction and possible reactivation ofjuxtaposed fault branches and segments that reflect thecomplex rupture termination. The aftershock distributionshowing the WNW-ESE alignment from the tomographicanalysis [Ayadi et al., 2008] and the existence of surfacecracks along the Thenia fault suggested the idea of a curvedfault model. Indeed, the bilateral rupture propagationobtained from body wave inversion [Delouis et al., 2004]and related maximum slip along the SW fault patch mayhave terminated against the Thenia fault and induced theN102�E trending surface cracks visible SW of Boumerdescity. The Thenia fault pointed out by the shift of fringes inFigures 4a and 4b is parallel to the aftershocks alignment as

Figure 11. Topographic and geologic sketch across the 2003 earthquake area. Red star is the mainshock [Bounif et al., 2004], and green dots are aftershocks [from Ayadi et al., 2008]. Tectonic structuresshow fold and nappe structures with south verging overthrusts and back thrusts [Meghraoui, 1988]. Thegeological units are made of Paleozoic basement rocks (G for granite and gneiss and Sch for micaschists)with intrusive basaltic units in red (Miocene), Mesozoic and Cenozoic substratum (Sb), Mesozoic andCenozoic nappes of flyschs (F), Triassic and Jurassic limestones (L), Postnappes Neogene sedimentarybasins (N), and Quaternary deposits (Q). The 2003 earthquake fault is well located by the main shock andrelated aftershocks, and its upward extension indicates the location (black arrow) of a possible fault scarpat �10 km distance from the epicenter in agreement with the modeled curved fault (see Figures 5b and6b), the seismic tomography [Ayadi et al., 2008], and the joint inversion of seismological waveforms andgeodetic data [Delouis et al., 2004; Yagi, 2003]. The gray arrow indicates the possible fault scarp locationif we adopt the flat and ramp fault geometry as interpreted from Deverchere et al. [2005].


14 of 16

B03406

clearly shown in the tomography of the western fault section[see Ayadi et al., 2008, Figures 2 and 3]. The Thenia faultand parallel fault branches can be considered as tear faultsthat accommodate the northwest thrust fault propagationwith regards to the Blida thrust system (Figure 1b). TheThenia fault branches may also act as a barrier and explainthe arrest of the lateral rupture propagation during the 2003earthquake.

5.4. Thrust Ruptures and Geometrical Complexities

[23] Other blind or hidden thrust fault structures showcomparable pattern of surface deformation and rupture geom-etry at depth with structural control of slip distribution andsize of fault patches [Stein and Yeats, 1989]. The 10 October1980 El Asnam earthquake (Ms 7.3) revealed �36-km-longthrust surface ruptures with three main segments strikingNE to ENE pointing out the complex rupture geometry atdepth of seismogenic faults in the Tell Atlas [Philip andMeghraoui, 1983; Chiarabba et al., 1997]. The 2 May 1983Coalinga earthquake (Ms 6.5) and related parallel faultbranches to the San Andreas fault took place on the 65�NE dipping reverse fault geometry inferred by the surfacedeformation and related surface folding [Stein and King,1984]. The 6 May 1976 Friuli earthquake (Ms 6.5, Alpinebelt in NE Italy) also shows complex thrust faulting withtwo �30� NNE dipping parallel patches with blind tosemiblind earthquake ruptures and a structural control ofthe fault propagation [Aoudia et al., 2000]. More recently,the subpixel analysis of SAR images of the 8 October 2005Kashmir earthquake (Mw 7.6) area indicates slip distributionon a �31� NE dipping thrust fault with three main segmentswhere the rupture initiation and arrest are on intersectingfaults [Pathier et al., 2006]. In our case, the aftershocksanalysis of the 2003 Zemmouri earthquake indicates incross sections en echelon SE dipping fault segments sepa-rated by a 3-km-wide step over [Ayadi et al., 2008]. Fromthe seismic moment tensor of main shock and aftershocks,Braunmiller and Bernardi [2005] suggest a NNE trendingtransform fault that limits the SW rupture. The focalmechanism solution of the 29 May 2008 (Mw 5.0) after-shock shows a high-angle NNE trending nodal fault planeand may be interpreted as a transform fault at a step overbetween the 2003 earthquake rupture farther NE and theBlida thrust system to the SW (Figure 1b). However, theinferred transform fault inland shows no evidence of InSARsignature comparable to that of the Thenia fault.

5.5. Seismic Hazard Implications

[24] Our study shows that an InSAR analysis along acoastal region is able to provide useful constraints on anearthquake rupture area and associated seismic hazard onneighboring fault segments (Figure 1b). Moreover, theMitidja basin displays to its northern edge the Sahelanticline that hides a 70-km-long NW dipping thrust faultin which the westernmost section was reactivated duringthe 1989 Tipaza earthquake (Mw 5.9 [Meghraoui, 1991])(Figure 1b). The fact that the two edges of the basin wererecently reactivated suggests that the central sections of eitherthe Sahel anticline or the Blida thrust fault system mustaccommodate shortening movements. The InSAR analysisand modeling of the 2003 Zemmouri earthquake ruptureallow us to constrain the SW rupture termination and may

contribute to a better seismic hazard evaluation. The possiblegeneration of a future Mw � 7 earthquake farther west alongthe nearby Blida or Sahel thrust fault segments that limit theMitidja basin (Figure 1b) put the Algiers city and populationat high seismic risk.

[25] Acknowledgments. This work was supported by the INSUresearch project ACI Cat-Nat ‘‘Risque sismique de la region d’Alger,’’and the Category 1 project 2532 and 2891 of the European Space Agency.RADARSAT-1 data (copyrighted by the Canadian Space Agency) wassupplied by the Alaskan Satellite Facility. Additional funding was obtainedfrom the U.S. Office of Foreign Disaster Assistance under the project on themodeling of seismic hazard of northern Algeria. Samir Belabbes wassupported by the Algerian Ministry of Higher Education and Research(MESRS) through a research studentship until December 2007 and morerecently by the Transfer project (**EC FP7, GOCE–CT-2006-37058);Ziyadin Cakir was supported by the Relief project (EC FP5 contractEVG1-2002-00069). We thank Ross Stein (USGS, Menlo Park), Jian Lin(WHOI, Boston), Tim Wright (Leeds University), Shinji Toda (AIST,Japan), Ahmet Akoglu (ITU, Turkey), Abdelhakim Ayadi (CRAAG,Algiers) and Catherine Dorbath (IPG Strasbourg) for comments anddiscussions on an early version of the manuscript. The bathymetry offshorethe earthquake area was kindly provided by Jacques Deverchere and AnneDomzig (Universite de Brest). We are grateful to Eric Fielding, SandySteacy, and an anonymous reviewer for their critical review of themanuscript. Some figures were prepared using the public domain GMTsoftware [Wessel and Smith, 1998].

ReferencesAkoglu, A. M., Z. Cakir, M. Meghraoui, S. Belabbes, S. O. El Alami,S. Ergintav, and S. Akyuz (2006), The 1994–2004 Al Hoceima (Morocco)earthquake sequence: Conjugate fault ruptures deduced from InSAR,Earth Planet. Sci. Lett., 252, 467–480, doi:10.1016/j.epsl.2006.10.010.

Alasset, P. J., H. Hebert, S. Maouche, V. Calbini, and M. Meghraoui (2006),The tsunami induced by the 2003 Zemmouri earthquake (Mw = 6.9,Algeria); modelling and results, Geophys. J. Int., 166, 213 – 226,doi:10.1111/j.1365-246X.2006.02912.x.

Aoudia, K., A. Sarao, B. Bukchin, and P. Suhadolc (2000), The 1976 Friuli(NE Italy) thrust earthquake: A reappraisal 23 years later, Geophys. Res.Lett., 27, 573–576, doi:10.1029/1999GL011071.

Ayadi, A., et al. (2003), Strong Algerian earthquake strikes near capital city,Eos Trans. AGU, 84, 561–568.

Ayadi, A., C. Dorbath, F. Ousadou, S. Maouche, M. Chikh, M. A. Bounif,and M. Meghraoui (2008), Zemmouri earthquake rupture zone (Mw 6.8,Algeria): Aftershocks sequence relocation and 3-D velocity model,J. Geophys. Res., 113, B09301, doi:10.1029/2007JB005257.

Belabbes, S., M. Meghraoui, Z. Cakir, and Y. Bouhadad (2008), InSARanalysis of a blind thrust rupture and related active folding: The 1999 AinTemouchent earthquake (Mw 5.7, Algeria) case study, J. Seismol.,doi:10.1007/s10950-008-9135-x.

Bounif, A., et al. (2004), The 21 May 2003 Zemmouri (Algeria) earthquakeMw 6.8; relocation and aftershock sequence analysis, Geophys. Res. Lett.,31, L19606, doi:10.1029/2004GL020586.

Braunmiller, J., and F. Bernardi (2005), The 2003 Boumerdes, Algeria,earthquake; regional moment tensor analysis, Geophys. Res. Lett., 32,L06305, doi:10.1029/2004GL022038.

Burgmann, R., P. Rosen, and E. Fielding (2000), Synthetic aperture radarinterferometry to measure Earth’s surface topography and its deformation,Annu. Rev. Earth Planet. Sci., 28, 169 – 209, doi:10.1146/annurev.earth.28.1.169.

Cakir, Z., M. Meghraoui, A. M. Akoglu, N. Jabour, S. Belabbes, and L. Ait-Brahim (2006), Surface deformation associated with the Mw 6.4, 24February 2004 Al Hoceima, Morocco, earthquake deduced from InSAR:Implications for the active tectonics along North Africa, Bull. Seismol.Soc. Am., 96, 59–68, doi:10.1785/0120050108.

Chiarabba, C., A. Amato, andM.Meghraoui (1997), Tomographic images ofthe El Asnam fault zone and the evolution of a seismogenic thrust-relatedfold, J. Geophys. Res., 102, 24,485–24,498, doi:10.1029/97JB01778.

Delouis, B., M. Vallee, M. Meghraoui, E. Calais, S. Maouche, K. Lamalli,A. Mahsas, P. Briole, F. Benhamouda, and A. Yelles (2004), Slip dis-tribution of the 2003 Boumerdes-Zemmouri earthquake, Algeria, fromteleseismic, GPS, and coastal uplift data, Geophys. Res. Lett., 31,L18607, doi:10.1029/2004GL020687.

Deverchere, J., et al. (2005), Active thrust faulting offshore Boumerdes,Algeria, and its relations to the 2003 Mw 6.9 earthquake, Geophys. Res.Lett., 32, L04311, doi:10.1029/2004GL021646.

Domzig, A., et al. (2006), Searching for the Africa-Eurasia Miocene bound-ary offshore western Algeria (MARADJA’03 cruise), in Quelques Devel-


15 of 16

B03406

oppements Recents sur la Geodynamique du Maghreb–Some RecentDevelopments on the Maghreb Geodynamics, pp. 80–91, Elsevier, Paris.

Farr, T. G., et al. (2007), The shuttle radar topography mission, Rev. Geo-phys., 45, RG2004, doi:10.1029/2005RG000183.

Fialko, Y., D. Sandwell, D. Agnew, M. Simons, P. Shearer, and B. Minster(2002), Deformation on nearby faults induced by the 1999 Hector Mineearthquake, Science, 297, 1858–1862, doi:10.1126/science.1074671.

Fielding, E. J., T. J. Wright, J. Muller, B. E. Parsons, and R. Walker (2004),Aseismic deformation of a fold-and-thrust belt imaged by synthetic aper-ture radar interferometry near Shahdad, southeast Iran, Geology, 32,577–580, doi:10.1130/G20452.1.

Goldstein, R. M., and C. L. Werner (1998), Radar interferogram filtering forgeophysical applications, Geophys. Res. Lett., 25(21), 4035–4038,doi:10.1029/1998GL900033.

Harbi, A., et al. (2007), Macroseismic study of the Zemmouri earthquake of21 May 2003 (Algeria Mw 6.8), Earthquake Spectra, 23, 315–332,doi:10.1193/1.2720363.

Lohman, R. B., and M. Simons (2005), Locations of selected small earth-quakes in the Zagros Mountains, Geochem. Geophys. Geosyst., 6,Q03001, doi:10.1029/2004GC000849.

Maerten, F., P. Resor, D. Pollard, and L. Maerten (2005), Inverting for slipon three-dimensional fault surfaces using angular dislocations, Bull. Seis-mol. Soc. Am., 95, 1654–1665, doi:10.1785/0120030181.

Mahsas, A., K. Lammali, K. Yelles, E. Calais, A. M. Freed, and P. Briole(2008), Shallow afterslip following the 2003May 21,Mw = 6.9 Boumerdesearthquake, Algeria, Geophys. J. Int., 172, 155–166, doi:10.1111/j.1365-246X.2007.03594.x.

Massonnet, D., and K. L. Feigl (1998), Radar interferometry and its appli-cation to changes in the Earth’s surface, Rev. Geophys., 36, 441–500,doi:10.1029/97RG03139.

Meghraoui, M. (1988), Geologie des zones sismiques du nord de l’Algerie:Paleosismologie, tectonique active et synthese sismotectonique, Ph.D.dissertation, Univ. Paris-Sud. Orsay, Orsay, France, 356 pp.

Meghraoui, M. (1991), Blind reverse faulting system associated with theMont Chenoua-Tipaza earthquake of 29 October 1989 (north-central Al-geria), Terra Nova, 3, 84–93, doi:10.1111/j.1365-3121.1991.tb00847.x.

Meghraoui, M., and F. Doumaz (1996), Earthquake-induced flooding andpaleoseismicity of the El Asnam, Algeria, fault-related fold, J. Geophys.Res., 101, 17,617–17,644, doi:10.1029/96JB00650.

Meghraoui, M., J. L. Morel, J. Andrieux, and M. Dahmani (1996), Tecto-nique plio-quaternaire de la chaine tello-rifaine et de la mer d’Alboran;une zone complexe de convergence continent-continent, Bull. Soc. Geol.Fr., 167, 141–157.

Meghraoui, M., S. Maouche, B. Chemaa, Z. Cakir, A. Aoudia, A. Harbi, P.-J.Alasset, A. Ayadi, Y. Bouhadad, and F. Benhamouda (2004), Coastal upliftand thrust faulting associated with theMw = 6.8 Zemmouri (Algeria) earth-quake of 21 May 2003, Geophys. Res. Lett., 31, L19605, doi:10.1029/2004GL020466.

Nocquet, J. M., and E. Calais (2004), Geodetic measurements of crustaldeformation in the western Mediterranean and Europe, Pure Appl. Geo-phys., 161, 661–681, doi:10.1007/s00024-003-2468-z.

Ouyed, M., M. Meghraoui, A. Cisternas, A. Deschamps, J. Dorel, J. Frechet,R. Gaulon, D. Hatzfeld, and H. Philip (1981), Seismotectonics of the ElAsnam earthquake, Nature, 292, 26–31, doi:10.1038/292026a0.

Pathier, E., E. Fielding, T. Wright, R. Walker, B. Parsons, and S. Hensley(2006), Displacement field and slip distribution of the 2005 Kashmirearthquake from SAR imagery, Geophys. Res. Lett., 33, L20310,doi:10.1029/2006GL027193.

Philip, H., and M. Meghraoui (1983), Structural analysis and interpretationof the surface deformations of the El Asnam earthquake of October 10,1980, Tectonics, 2, 17–49, doi:10.1029/TC002i001p00017.

Resor, P. G., D. D. Pollard, T. J. Wright, and G. C. Beroza (2005), Integrat-ing high-precision aftershock locations and geodetic observations tomodel coseismic deformation associated with the 1995 Kozani-Grevenaearthquake, Greece, J. Geophys. Res., 110, B09402, doi:10.1029/2004JB003263.

Scharroo, R., and P. Visser (1998), Precise orbit determination and gravityfield improvement for the ERS satellites, J. Geophys. Res., 103, 8113–8127, doi:10.1029/97JC03179.

Semmane, F., M. Campillo, and F. Cotton (2005), Fault location and sourceprocess of the Boumerdes, Algeria, earthquake inferred from geodeticand strong motion data, Geophys. Res. Lett., 32, L01305, doi:10.1029/2004GL021268.

Serpelloni, E., G. Vannucci, S. Pondrelli, A. Argnani, G. Casula, M. Anzidei,P. Baldi, and P. Gasperini (2007), Kinematics of the western Africa-Eurasiaplate boundary from focal mechanisms and GPS data, Geophys. J. Int.,169, 1180–1200, doi:10.1111/j.1365-246X.2007.03367.x.

Stein, R., and G. King (1984), Seismic potential revealed by surface fold-ing: 1983 Coalinga, California, earthquake, Science, 224, 869–872,doi:10.1126/science.224.4651.869.

Stein, R., and R. Yeats (1989), Hidden earthquakes: Large earthquakes cantake place not only on faults that cut the Earth’s surface but also on‘‘blind’’ faults under folded terrain, Sci. Am., 260(6), 48–57.

Stich, D., C. J. Ammon, and J. Morales (2003), Moment tensor solutionsfor small and moderate earthquakes in the Ibero-Maghreb region,J. Geophys. Res., 108(B3), 2148, doi:10.1029/2002JB002057.

Thomas, A. L. (1993), A three-dimensional, polygonal element, displace-ment discontinuity boundary element computer with applications to frac-tures, fault, and cavities in the Earth’s crust, Master’s thesis, 221 pp.,Stanford Univ., Stanford, Calif.

Wessel, P., and H. F. Smith (1998), New, improved version of the GenericMapping Tools released, Eos Trans. AGU, 79, 579, doi:10.1029/98EO00426.

Wright, T. J., Z. Lu, and C. Wicks (2004), Constraining the slip distributionand fault geometry of the Mw 7.9, 3 November 2002, Denali fault earth-quake with interferometric synthetic aperture radar and Global Position-ing System data, Bull. Seismol. Soc. Am., 94, S175–S189, doi:10.1785/0120040623.

Yagi, Y. (2003), Source process of large and significant earthquakes in2003, Bull. Int. Inst. Seismol. Earthquake Eng., special volume, 145–153.

Yelles, K., K. Lammali, A. Mahsas, E. Calais, and P. Briole (2004), Co-seismic deformation of the May 21st, 2003 earthquake, Algeria, fromGPS measurements, Geophys. Res. Lett., 31, L13610, doi:10.1029/2004GL019884.

Yielding, G., M. Ouyed, G. C. P. King, and D. Hatzfeld (1989), Activetectonics of the Algerian Atlas mountains—Evidence from aftershocks ofthe 1980 El Asnam earthquake, Geophys. J. Int., 99(3), 761 –788,doi:10.1111/j.1365-246X.1989.tb02057.x.

Zebker, H. A., J. F. Vesecky, and Q. Lin (1994), Phase wrapping throughfringe-line detection in synthetic aperture radar interferometry, Appl.Opt., 33, 201–208.

��S. Belabbes and M. Meghraoui, Institut de Physique du Globe de

Strasbourg, UMR 7516, 5, rue Rene Descartes, F-670084 Strasbourg,France. ([email protected])Z. Cakir, Faculty of Mines, Istanbul Technical University, 34469 Maslak,

Istanbul, Turkey.C. Wicks, U.S. Geological Survey, 345 Middlefield Road, MS 977,

Menlo Park, CA 94025, USA.


16 of 16

B03406

rupture parameters of the 2003 zemmouri (m 6.8), algeria...

Documents